feat(integrations): add ApiSmart OpenAI-compatible gateway provider

README.md

@@ -268,6 +268,7 @@ Advanced and source-build guides:
 | OpenAI-compatible | `/provider` or env vars | Works with OpenAI, OpenRouter, DeepSeek, Groq, Mistral, LM Studio, and other compatible `/v1` servers |
 | Z.AI GLM Coding Plan | `/provider` or OpenAI-compatible env vars | Uses `OPENAI_API_KEY` at `https://api.z.ai/api/coding/paas/v4` and defaults to `glm-5.2` |
 | AI/ML API | `/provider` or `AIMLAPI_API_KEY` ([setup guide](docs/aimlapi-setup.md)) | Uses `https://api.aimlapi.com/v1`, auto-detects the OpenAI-compatible route from `AIMLAPI_API_KEY`, sends OpenClaude attribution headers, and discovers chat-capable models from the public `/models` catalog |
+| ApiSmart | `/provider` or `APISMART_API_KEY` | Uses `https://gw.apismart.ai/v1`, defaults to `DEEPSEEK_V4_FLASH`, and supports optional `APISMART_MODEL` plus authenticated model discovery |
 | Hicap | `/provider` or OpenAI-compatible env vars | Uses `api-key` auth, discovers models from unauthenticated `/models`, and supports Responses mode for `gpt-` models |
 | Fireworks AI | `/provider` or env vars | First-class provider with 276 curated models (DeepSeek, Qwen, Llama, Gemma, and more); uses `FIREWORKS_API_KEY` |
 | LongCat | `/provider` or env vars | Meituan LongCat OpenAI-compatible API at `https://api.longcat.chat/openai/v1`; uses `LONGCAT_API_KEY` and defaults to `LongCat-2.0` |

scripts/system-check.test.ts

@@ -82,6 +82,8 @@ const ENV_KEYS = [
   'VENICE_API_KEY',
   'MIMO_API_KEY',
   'ATLAS_CLOUD_API_KEY',
+  'APISMART_API_KEY',
+  'APISMART_MODEL',
   'NEARAI_API_KEY',
   'FIREWORKS_API_KEY',
   'CLINE_API_KEY',

src/integrations/compatibility.test.ts

@@ -19,6 +19,7 @@ import type { ProviderPreset } from '../utils/providerProfiles.js'
 const EXPECTED_PRESETS = [
   'anthropic',
   'atlas-cloud',
+  'apismart',
   'aimlapi',
   'openai',
   'ollama',

src/integrations/discoveryService.test.ts

@@ -18,6 +18,7 @@ const originalEnv = {
   OPENAI_API_KEY: process.env.OPENAI_API_KEY,
   OPENAI_API_KEYS: process.env.OPENAI_API_KEYS,
   OPENAI_MODEL: process.env.OPENAI_MODEL,
+  APISMART_API_KEY: process.env.APISMART_API_KEY,
   ANTHROPIC_CUSTOM_HEADERS: process.env.ANTHROPIC_CUSTOM_HEADERS,
   CLAUDE_CODE_USE_OPENAI: process.env.CLAUDE_CODE_USE_OPENAI,
   CLAUDE_CODE_USE_GEMINI: process.env.CLAUDE_CODE_USE_GEMINI,
@@ -59,6 +60,7 @@ function clearProviderEnv(): void {
   delete process.env.OPENAI_API_KEY
   delete process.env.OPENAI_API_KEYS
   delete process.env.OPENAI_MODEL
+  delete process.env.APISMART_API_KEY
   delete process.env.ANTHROPIC_CUSTOM_HEADERS
   delete process.env.CLAUDE_CODE_USE_OPENAI
   delete process.env.CLAUDE_CODE_USE_GEMINI
@@ -93,6 +95,7 @@ afterEach(() => {
     restoreEnvValue('OPENAI_API_KEY')
     restoreEnvValue('OPENAI_API_KEYS')
     restoreEnvValue('OPENAI_MODEL')
+    restoreEnvValue('APISMART_API_KEY')
     restoreEnvValue('ANTHROPIC_CUSTOM_HEADERS')
     restoreEnvValue('CLAUDE_CODE_USE_OPENAI')
     restoreEnvValue('CLAUDE_CODE_USE_GEMINI')
@@ -110,6 +113,30 @@ afterEach(() => {
 })
 
 describe('discoverModelsForRoute', () => {
+  test('does not send an ApiSmart key to an overridden discovery URL', async () => {
+    const { discoverModelsForRoute } = await loadDiscoveryServiceModule()
+    process.env.APISMART_API_KEY = 'apismart-secret'
+    let didFetch = false
+    let authorization: string | null | undefined
+    setMockFetch(mock((_input: string | URL | Request, init?: RequestInit) => {
+      didFetch = true
+      authorization = new Headers(init?.headers).get('authorization')
+      return Promise.resolve(
+        new Response(JSON.stringify({ data: [] }), {
+          headers: { 'Content-Type': 'application/json' },
+        }),
+      )
+    }) as unknown as typeof globalThis.fetch)
+
+    await discoverModelsForRoute('apismart', {
+      baseUrl: 'https://proxy.example/v1',
+      forceRefresh: true,
+    })
+
+    expect(didFetch).toBe(true)
+    expect(authorization).not.toBe('Bearer apismart-secret')
+  })
+
   test('uses built-in openai-compatible discovery and caches results for dynamic routes', async () => {
     const { discoverModelsForRoute } = await loadDiscoveryServiceModule()
 

src/integrations/discoveryService.ts

@@ -15,6 +15,7 @@ import type {
 import { resolveRouteIdFromBaseUrl } from './index.js'
 import {
   getRouteDescriptor,
+  isCanonicalApismartInferenceBaseUrl,
   resolveActiveRouteIdFromEnv,
   resolveRouteCredentialValue,
 } from './routeMetadata.js'
@@ -152,12 +153,23 @@ function getRouteBaseUrl(
 
 function getRouteDiscoveryApiKey(
   routeId: string,
-  options?: { apiKey?: string },
+  options?: { baseUrl?: string; apiKey?: string },
 ): string | undefined {
   if (getRouteCatalog(routeId)?.discovery?.requiresAuth === false) {
     return undefined
   }
 
+  const baseUrl = getRouteBaseUrl(routeId, options)
+  // ApiSmart's dedicated token must never be used for an overridden discovery
+  // URL. Apply the same exact inference-endpoint boundary used by requests and
+  // profiles before considering either a caller-provided or ambient key.
+  if (
+    routeId === 'apismart' &&
+    !isCanonicalApismartInferenceBaseUrl(baseUrl)
+  ) {
+    return undefined
+  }
+
   if (hasInvalidCredentialPlaceholder(options?.apiKey)) {
     return undefined
   }
@@ -170,6 +182,7 @@ function getRouteDiscoveryApiKey(
   return firstUsableCredential(
     resolveRouteCredentialValue({
       routeId,
+      baseUrl,
       processEnv: process.env,
     }),
   )

src/integrations/gateways/apismart.test.ts

@@ -0,0 +1,44 @@
+import { describe, expect, test } from 'bun:test'
+
+import apismart from './apismart.js'
+
+const mapModel = apismart.catalog?.discovery?.mapModel
+
+describe('apismart gateway', () => {
+  test('uses hybrid discovery with dedicated credentials', () => {
+    expect(apismart.id).toBe('apismart')
+    expect(apismart.catalog?.source).toBe('hybrid')
+    expect(apismart.catalog?.discovery?.kind).toBe('openai-compatible')
+    expect(apismart.catalog?.discovery?.requiresAuth).toBe(true)
+    expect(apismart.setup.dedicatedCredentialsOnly).toBe(true)
+    expect(apismart.setup.credentialEnvVars).toEqual(['APISMART_API_KEY'])
+    expect(apismart.defaultBaseUrl).toBe('https://gw.apismart.ai/v1')
+    expect(apismart.defaultModel).toBe('DEEPSEEK_V4_FLASH')
+    expect(mapModel).toBeDefined()
+  })
+
+  test('curated catalog keeps ApiSmart case-sensitive model ids', () => {
+    const models = apismart.catalog?.models ?? []
+    expect(models.some(model => model.apiName === 'DEEPSEEK_V4_FLASH')).toBe(
+      true,
+    )
+    expect(models.some(model => model.apiName === 'GLM_5.2')).toBe(true)
+    expect(models.some(model => model.apiName === 'QWEN_3_7_MAX')).toBe(true)
+  })
+
+  test('mapModel keeps chat models and drops image/video ids', () => {
+    if (!mapModel) throw new Error('mapModel missing')
+    expect(mapModel({ id: 'DEEPSEEK_V4_FLASH', owned_by: 'DeepSeek' })).toEqual(
+      {
+        id: 'DEEPSEEK_V4_FLASH',
+        apiName: 'DEEPSEEK_V4_FLASH',
+        label: 'DEEPSEEK_V4_FLASH (DeepSeek)',
+      },
+    )
+    expect(mapModel({ id: 'seedream-5.0' })).toBeNull()
+    expect(mapModel({ id: 'seedance-2.0' })).toBeNull()
+    expect(mapModel({ id: 'happyhorse-1.0' })).toBeNull()
+    expect(mapModel({})).toBeNull()
+    expect(mapModel(null)).toBeNull()
+  })
+})

src/integrations/gateways/apismart.ts

@@ -0,0 +1,227 @@
+import { defineGateway } from '../define.js'
+
+const NON_CHAT_MODEL_PATTERN =
+  /(seedream|seedance|happyhorse|image|video|vision-preview)/i
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === 'object' && value !== null
+}
+
+function getTrimmedString(
+  record: Record<string, unknown>,
+  key: string,
+): string | undefined {
+  const value = record[key]
+  return typeof value === 'string' ? value.trim() : undefined
+}
+
+function mapApismartModel(raw: unknown) {
+  if (!isRecord(raw)) {
+    return null
+  }
+
+  const id = getTrimmedString(raw, 'id')
+  if (!id || NON_CHAT_MODEL_PATTERN.test(id)) {
+    return null
+  }
+
+  const ownedBy =
+    getTrimmedString(raw, 'owned_by') || getTrimmedString(raw, 'ownedBy')
+  const label = ownedBy ? `${id} (${ownedBy})` : id
+
+  return {
+    id,
+    apiName: id,
+    label,
+  }
+}
+
+// Curated from https://www.apismart.ai/models (PAGE_MODEL_LIST LLM entries).
+// Exact Model IDs are case-sensitive per ApiSmart docs.
+const curatedModels = [
+  {
+    id: 'DEEPSEEK_V4_FLASH',
+    apiName: 'DEEPSEEK_V4_FLASH',
+    aliases: ['deepseek-v4-flash'],
+    modelDescriptorId: 'deepseek-v4-flash',
+    label: 'DeepSeek V4 Flash',
+    contextWindow: 1_048_576,
+    maxOutputTokens: 8_192,
+  },
+  {
+    id: 'DEEPSEEK_V4_PRO',
+    apiName: 'DEEPSEEK_V4_PRO',
+    aliases: ['deepseek-v4-pro'],
+    modelDescriptorId: 'deepseek-v4-pro',
+    label: 'DeepSeek V4 Pro',
+    contextWindow: 1_048_576,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'DEEPSEEK_V3_2',
+    apiName: 'DEEPSEEK_V3_2',
+    aliases: ['deepseek-v3.2'],
+    modelDescriptorId: 'deepseek-ai/deepseek-v3.2',
+    label: 'DeepSeek V3.2',
+    contextWindow: 131_072,
+    maxOutputTokens: 8_192,
+  },
+  {
+    id: 'KIMI_K2_6',
+    apiName: 'KIMI_K2_6',
+    aliases: ['kimi-k2.6'],
+    modelDescriptorId: 'kimi-k2.6',
+    label: 'Kimi K2.6',
+    contextWindow: 262_144,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'KIMI_K2_5',
+    apiName: 'KIMI_K2_5',
+    aliases: ['kimi-k2.5'],
+    modelDescriptorId: 'kimi-k2.5',
+    label: 'Kimi K2.5',
+    contextWindow: 262_144,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'KIMI_K3',
+    apiName: 'KIMI_K3',
+    aliases: ['kimi-k3', 'k3'],
+    modelDescriptorId: 'k3',
+    label: 'Kimi K3',
+    contextWindow: 1_075_200,
+    maxOutputTokens: 8_192,
+  },
+  {
+    id: 'GLM_5.2',
+    apiName: 'GLM_5.2',
+    aliases: ['glm-5.2'],
+    modelDescriptorId: 'glm-5.2',
+    label: 'GLM 5.2',
+    contextWindow: 1_048_576,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'GLM_5_1',
+    apiName: 'GLM_5_1',
+    aliases: ['glm-5.1'],
+    modelDescriptorId: 'glm-5.1',
+    label: 'GLM 5.1',
+    contextWindow: 204_800,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'GLM_5',
+    apiName: 'GLM_5',
+    aliases: ['glm-5'],
+    modelDescriptorId: 'glm-5',
+    label: 'GLM 5',
+    contextWindow: 209_920,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'QWEN_3_7_MAX',
+    apiName: 'QWEN_3_7_MAX',
+    aliases: ['qwen3.7-max'],
+    modelDescriptorId: 'qwen3.7-max',
+    label: 'Qwen 3.7 Max',
+    contextWindow: 1_048_576,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'QWEN_3_6_PLUS',
+    apiName: 'QWEN_3_6_PLUS',
+    aliases: ['qwen3.6-plus'],
+    modelDescriptorId: 'qwen3.6-plus',
+    label: 'Qwen 3.6 Plus',
+    contextWindow: 1_048_576,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'MINIMAX_M3',
+    apiName: 'MINIMAX_M3',
+    aliases: ['minimax-m3'],
+    modelDescriptorId: 'minimax-m3',
+    label: 'MiniMax M3',
+    contextWindow: 1_048_576,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'MINIMAX_M2_5',
+    apiName: 'MINIMAX_M2_5',
+    aliases: ['minimax-m2.5'],
+    modelDescriptorId: 'minimax-m2.5',
+    label: 'MiniMax M2.5',
+    contextWindow: 200_704,
+    maxOutputTokens: 16_384,
+  },
+  {
+    id: 'MINIMAX_M2_5_HIGHSPEED',
+    apiName: 'MINIMAX_M2_5_HIGHSPEED',
+    aliases: ['minimax-m2.5-highspeed'],
+    modelDescriptorId: 'minimax-m2.5-highspeed',
+    label: 'MiniMax M2.5 HighSpeed',
+    contextWindow: 204_800,
+    maxOutputTokens: 16_384,
+  },
+]
+
+export default defineGateway({
+  id: 'apismart',
+  label: 'ApiSmart',
+  category: 'aggregating',
+  defaultBaseUrl: 'https://gw.apismart.ai/v1',
+  defaultModel: 'DEEPSEEK_V4_FLASH',
+  supportsModelRouting: true,
+  setup: {
+    requiresAuth: true,
+    authMode: 'api-key',
+    credentialEnvVars: ['APISMART_API_KEY'],
+    dedicatedCredentialsOnly: true,
+  },
+  startup: {
+    probeReadiness: 'openai-compatible-models',
+  },
+  transportConfig: {
+    kind: 'openai-compatible',
+    openaiShim: {
+      supportsApiFormatSelection: false,
+      supportsAuthHeaders: false,
+      // ApiSmart chat-completions examples use max_tokens.
+      maxTokensField: 'max_tokens',
+    },
+  },
+  preset: {
+    id: 'apismart',
+    description: 'ApiSmart unified OpenAI-compatible gateway',
+    vendorId: 'openai',
+    apiKeyEnvVars: ['APISMART_API_KEY'],
+    modelEnvVars: ['APISMART_MODEL', 'OPENAI_MODEL'],
+  },
+  validation: {
+    kind: 'credential-env',
+    routing: {
+      matchDefaultBaseUrl: true,
+      matchBaseUrlHosts: ['gw.apismart.ai'],
+    },
+    credentialEnvVars: ['APISMART_API_KEY'],
+    missingCredentialMessage:
+      'ApiSmart auth is required. Set APISMART_API_KEY.',
+  },
+  catalog: {
+    // /v1/models exists and requires a Bearer key; curated LLM ids stay visible
+    // before discovery and when refresh fails.
+    source: 'hybrid',
+    discovery: {
+      kind: 'openai-compatible',
+      requiresAuth: true,
+      mapModel: mapApismartModel,
+    },
+    discoveryCacheTtl: '1d',
+    discoveryRefreshMode: 'background-if-stale',
+    allowManualRefresh: true,
+    models: [...curatedModels],
+  },
+  usage: { supported: false },
+})